Mid-Season Updates to Rim Protection and True Usage Stats (SportVU aided)
As you might have noticed, NBA.COM now includes (some) SportVU player tracking data in each game's box score (e.g.). I'll have a longer piece soon on using and interpreting this box score level data. But, now that we're at the All-Star Break, now seems like a natural time to update the Isolated Rim Protection and TrueUsage/TrueTurnover Rate metrics I've been looking at all season.
I'm still really happy with how the Rim Protection stuff is working (especially with the assistance of Darryl's data tabulating the number of shots opponents attempt within 5 feet while each player is on the floor). I'm very much interested in getting multiple seasons' worth of data just to see how much of what the SportVU cameras are picking up is scheme, and how much is individual player ability. If he ever gets on the court in Indy, Andrew Bynum will be an interesting test, for example. One of the continuing surprises of the season is how well Ian Mahinmi has shown up in this data - he hasn't quite matched Roy Hibbert's rim protection output, but has been pretty close. However, the Pacers are about 12 points per 100 possessions worse on defense with Mahinmi in and Hibbert out than the reverse.
Obviously there are confounding factors (Luis Scola is not David West defensively, Hibbert spends a greater proportion of his time on the floor with Paul George than does Mahinmi), and rim protection is only a slice of defensive value. But, Mahinmi's numbers being so close to Hibbert's on this metric make me wonder how much of this is simply Indiana's ability to scheme defensively to make sure their 5-man is always protecting the rim and allowing West and Scola to chase around the perimeter on pick-and-roll defense. Player movement within this season and more seasons' worth of data are probably necessary to parse this out.
Anyway, the updated doc with the Rim Protection values in terms of "adjusted points saved" is here (short version the points "saved" by that player's presence near the rim, adjusted for number of contesting opportunities relative to league average - basically guys who are on the court for lower numbers of opponent attempts from <5 ft get a bump, higher number of attempts allowed means a slight penalty.)
Obviously there are confounding factors (Luis Scola is not David West defensively, Hibbert spends a greater proportion of his time on the floor with Paul George than does Mahinmi), and rim protection is only a slice of defensive value. But, Mahinmi's numbers being so close to Hibbert's on this metric make me wonder how much of this is simply Indiana's ability to scheme defensively to make sure their 5-man is always protecting the rim and allowing West and Scola to chase around the perimeter on pick-and-roll defense. Player movement within this season and more seasons' worth of data are probably necessary to parse this out.
Anyway, the updated doc with the Rim Protection values in terms of "adjusted points saved" is here (short version the points "saved" by that player's presence near the rim, adjusted for number of contesting opportunities relative to league average - basically guys who are on the court for lower numbers of opponent attempts from <5 ft get a bump, higher number of attempts allowed means a slight penalty.)
The other metric I've been working with is an attempt to re-imagine measurements of players' offensive roles. I go into it in more detail here and linked posts, but essentially I'm treating each possession as potentially collaborative as opposed to necessarily rivalrous. It's possible for 2 and sometimes even 3 offensive players to "use" a given possession, which I think aligns somewhat better to how basketball is actually played than the "1 player to a possession" assumption inherent in traditional usage and turnover rates.
This metric also allows for a more accurate estimation of each player's propensity for turnovers relative to "doing stuff." Traditional TOV% is a stat which on first pass "rewards" volume shooting since the denominator is basically shot attempts. Pass-first point guards have always looked like "high turnover" players under this approach, which I don't think is really accurate - these players handle the ball a large proportion of the time, and in general to be an NBA-level point guard, you actually turn the ball over extremely rarely, relatively speaking. So, the key stats to look at are "True Usage" which expresses a percentage of a team's possessions a player either shoots, turns the ball over, generates and assist chance or is credited with a "hockey assist" while on the floor. TrueTurnover% measures the ratio of these "play ending events" which are turnovers. That doc is available here.
For T'Wolves fans, not the JJ Barea has the 12th highest "True Usage" in the league, which I think explains a lot about a lot. (See also this breakdown of Barea's end of quarter struggles)
This metric also allows for a more accurate estimation of each player's propensity for turnovers relative to "doing stuff." Traditional TOV% is a stat which on first pass "rewards" volume shooting since the denominator is basically shot attempts. Pass-first point guards have always looked like "high turnover" players under this approach, which I don't think is really accurate - these players handle the ball a large proportion of the time, and in general to be an NBA-level point guard, you actually turn the ball over extremely rarely, relatively speaking. So, the key stats to look at are "True Usage" which expresses a percentage of a team's possessions a player either shoots, turns the ball over, generates and assist chance or is credited with a "hockey assist" while on the floor. TrueTurnover% measures the ratio of these "play ending events" which are turnovers. That doc is available here.
For T'Wolves fans, not the JJ Barea has the 12th highest "True Usage" in the league, which I think explains a lot about a lot. (See also this breakdown of Barea's end of quarter struggles)
RSS Feed